- Email: [email protected]
Autonomic changes during wake–sleep transition: A heart rate variability based approach
Autonomic Neuroscience: Basic and Clinical 130 (2006) 17 – 27 www.elsevier.com/locate/autneu
Autonomic changes during wake–sleep transition: A heart rate variability based approach Zvi Shinar a,⁎, Solange Akselrod a , Yaron Dagan b , Armanda Baharav a a b
Abramson Center for Medical Physics, Tel Aviv University, Tel-Aviv, Israel Sleep and Fatigue Institute, SHEBA Medical Center, Tel Hashomer, Israel
Received 8 November 2005; received in revised form 3 April 2006; accepted 28 April 2006
Abstract Autonomic function during sleep and wakefulness has been extensively investigated, however information concerning autonomic changes during the wake to sleep transition is scarce. The objective of the present study was to non-invasively characterize autonomic function and additional physiologic changes during sleep onset in normal and abnormal sleep. The estimation of autonomic function was based on time– frequency analysis of the RR interval series, using the power components in the very-low-frequency range (0.005–0.04 Hz), low-frequency (0.04–0.15 Hz), and high-frequency range (0.15–0.5 Hz). The ratio of low to high frequency power represented the sympathovagal balance. Thirty-four subjects who underwent whole night polysomnography were divided into 3 groups according to their complaints and study results: normal subjects, apneic patients (OSAS), and subjects with various sleep disorders (VSD). The results indicated a significant increase in RR interval during sleep onset, although its variability decreased; respiratory rate did not change, yet respiration became more stable; EMG amplitude and its variability decreased with sleep onset. Very-low-frequency power started to decrease significantly 2 min before sleep onset in all groups; low-frequency power decreased and high-frequency power did not change significantly in all groups, accordingly their ratio decreased and reflected a shift towards parasympathetic predominance. Although autonomic function displayed similar behavior in all subjects, OSAS and VSD patients presented a higher sympathovagal balance reflecting enhanced sympathetic predominance in those groups compared to normal subjects, both before and after sleep onset. All parameters reached a nadir at a defined time point during the process of falling asleep. We conclude that the wake–sleep transition period represents a transitional process between two physiologically different states; this transition starts with a decrease in the very slow oscillations in heart rate that anticipates a step-change resetting of autonomic function, followed by a decrease in sympathovagal balance towards the end of the process. © 2006 Elsevier B.V. All rights reserved. Keywords: Autonomic nervous system; HRV; Wavelet; Sleep onset; Electrocardiogram
1. Introduction Sleep Onset (SO) is a transition process between two different physiological states: wakefulness and sleep, usually Non Rapid Eye Movement (NREM) sleep. The central regulation of autonomic function differs greatly between NREM and Rapid Eye Movement (REM) sleep (Parmeg⁎ Corresponding author. Medical Physics, School of Physics and Astronomy, Tel Aviv University, Tel Aviv 69978, Israel. Tel.: +972 3 6408669 or 6408204; fax: +972 3 6406237. E-mail address: [email protected] (Z. Shinar). 1566-0702/$ - see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.autneu.2006.04.006
giani and Morrison, 1990; Somers et al., 1993), yet little is known about the changes that occur during SO specifically. The definition of SO is rather vague as it depends on the physiologic or behavioral parameter under specific scrutiny (Ogilvie, 2001). Thus the standard definition (Rechtschaffen and Kales, 1968) of SO is based on consistent and reproducible changes in EEG, EOG, and EMG, and it usually allows to detect the transition from wakefulness to sleep within seconds. The subjective perception of falling asleep does not always coincide with the timing of the event as estimated by the standard definition. The entire process of falling asleep is further complicated by the fact that SO does not occur all at
18
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27
once (Ogilvie, 2001; Ogilvie et al., 1989) and some fluctuations in vigilance may occur before stable, unequivocal sleep develops. The complexity of the normal transition from quiet wakefulness to NREM sleep can be further demonstrated by the study of behavioral parameters during SO, such as the performance of simple behavioral tasks, memory tasks, or the response to auditory or visual stimuli (Bonnet and Moore, 1982; Foulkes and Vogel, 1965; Vogel et al., 1966). Alpha frequency EEG is typical for wakefulness with closed eyes, and in most subjects it appears before SO. The power in Delta frequency band increases during light sleep (NREM stage 2) as an intermediate state between wake and deep sleep. Delta frequency EEG is dominant during the deep stages of NREM sleep. The direct investigation of autonomic nervous function is invasive, leading by itself to unwanted changes and reactions. Thus direct studies of autonomic activity are usually avoided due to their relative complexity and invasiveness. Fortunately enough, important information can be inferred from the response of end organs in general, and from the heart rate variability (HRV) in particular. Thus standard methods of HRV in the time domain (Fast Fourier Transform or Autoregression) are widely accepted non-invasive tools, that allow to obtain important information about central autonomic regulation (Akselrod, 1995; Berntson et al., 1997;Malliani, 1995). The obtained HRV-power spectrum typically includes: a low frequency (LF) power component between (0.04– 0.15 Hz) which (in supine position) reflects both sympathetic and parasympathetic effects on the sinus node, and a distinct high frequency (HF) power component, around the respiratory frequency (0.15–0.5 Hz) which is mediated mainly by parasympathetic activity. Since LF includes information about both branches of the ANS, an estimate of autonomic or sympathovagal balance has been defined as the LF/HF ratio (Malliani, 1995). Slower HRV fluctuations in the range (0.005–0.04 Hz) known as Very Low Frequency (VLF) have been attributed to vasomotion, thermoregulation, and renin– angiotensin effects (Akselrod, 1995; Fleisher et al., 1996; Shefi, 1997; Taylor et al., 1998), and may be influenced, at least in part, by mental or physical activity (Bernardi et al., 1996, 2000). Thus VLF, LF, HF, and the LF/HF ratio provide a reliable estimate of cardiac autonomic control. One of the main disadvantages of power spectrum analysis resides in its limitation to steady state conditions. Timedependent spectral decomposition of beat-to-beat variability in HR overcomes this stationarity limitation (Keselbrener and Akselrod, 1996; Toledo et al., 2003), and can be applied in non-steady state conditions encountered frequently in physiologic studies. This approach is necessary for the evaluation of the transient autonomic changes that accompany SO. Several direct measurements (Hornyak et al., 1991) based on muscle sympathetic nerve activity and many indirect (Baharav et al., 1995; Berlad et al., 1993; Van de Borne et al., 1994) studies based on instantaneous heart rate variability as an estimate of autonomic activity addressed ANS function during sleep. These studies indicated that parasympathetic
activity increases progressively after SO and sympathetic nervous system activity decreases after falling asleep. These trends continued with deepening NREM sleep. However, only a small number of studies dealt specifically with ANS function during SO. Most of these studies were based on a series of experiments specifically designed to identify circadian effects (Burgess et al., 1997, 1999a,b; Carrington et al., 2003) concerning the influence of sleep on measures of cardiac autonomic activity. They indicated that blood pressure decreased and sympathovagal balance shifted towards increased vagal activity, in close association with SO. Moreover respiratory sinus arrhythmia, as a measure of parasympathetic activity, showed a 24-h rhythm independent of sleep, whereas pre-ejection period as a measure of sympathetic activity only showed a 24-h rhythm if sleep occurred. These findings suggest that parasympathetic nervous system activity was under strong circadian influence, whereas the sympathetic nervous system activity was mostly influenced by the sleep system (Burgess et al., 1997). More recent results indicated that the shift towards vagal dominance was primarily a function of sleep, whereas the change in sympathetic activity at sleep onset reflected circadian influence (Carrington et al., 2003). Although the results of different studies aimed to uncover circadian and sleep related influences on autonomic function around wake–sleep transitions are not in total agreement, there is no doubt that important changes in autonomic activity occur around sleep onset. The objective of the present study was to characterize the wake–sleep transition process, with the main focus on autonomic function changes as reflected by the instantaneous HRV. We estimated the behavior of additional physiologic variables (HR, respiration, muscle activity, and cerebral activity as reflected in the electroencephalogram) around SO. We assumed that falling asleep presents a consistent pattern of changes in these variables, that are common to normal and abnormal sleep. 2. Materials and methods 2.1. Data acquisition Thirty-four patients referred to a sleep laboratory underwent whole night sleep studies. All subjects gave their informed consent to participate in the study and answered a sleep questionnaire. Studies included whole night polysomnography with digitization and recording of the following channels: 4 EEG channels (two central and two occipital leads), and 2 EOG channels sampled at 100 Hz with low pass filter at 35 Hz, chin EMG, and right/left Anterior Tibialis (EMG)-sampled at 100 Hz low pass filtered at 45 Hz; ECG channel sampled at 200 Hz, with low pass at 100 Hz and notch filter at 50 Hz; abdominal and chest respiratory effort, nasal and oral airflow, end tidal CO2, oxygen saturation and pulse wave, all sampled at 10 Hz, and low pass filtered at 5 Hz. Subjects were asked to lie down 5 min before lights off.
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27
The polysomnograms were manually scored, by an expert, according to Rechtschaffen and Kales (R&K) criteria (Rechtschaffen and Kales, 1968), and were screened for any abnormalities. Following the scoring procedure, subjects were classified in 3 groups: 1. Normal patients (Normals)—No sleep abnormality was diagnosed after the sleep study in otherwise healthy subjects referred to the sleep clinic for mild snoring or some vague sleep quality or daytime fatigue complaints— 12 subjects (6 males): average ± Standard Deviation for age: 28 ± 16 years, sleep latency: 11 ± 8 min, sleep efficiency: 89 ± 12%. 2. Patients with obstructive sleep apnea (OSAS)—11 subjects (8 males): average ± Standard Deviation for age: 40 ± 9 years, sleep latency: 13 ± 9 min, and sleep efficiency: 86 ± 9%, average RDI 33.3 ± 19.6. 3. Patients with various sleep disorders (VSD)—11 subjects (8 males): average ± Standard Deviation for age: 37 ± 18 years, sleep latency: 10 ± 7 min, and sleep efficiency: 87 ± 8%. Subjects in this group were referred to the sleep laboratory with a main complaint of fatigue. Seven of them had a background of a psychiatric disease (obsessive– compulsive disorder, or post-traumatic stress disorder), two a history of asthma, and the other two were otherwise healthy; however their sleep study revealed periodic leg movement syndrome in one (PLM index 6.4), and central apnea (RDI 28.8) in the other. None had OSAS. No subject received any direct sympathetic agonist therapy, one patient in the OSAS group and two in the VSD received beta-blockers. 2.2. Data analysis 2.2.1. EEG signals The power of the EEG in the frequency bands of Delta (0.5–4 Hz), Theta (4–7 Hz), Alpha (7.5–14), and Beta (14– 35 Hz) was calculated for segments of 6 s. The power in each
19
frequency band was normalized by the total power in each segment. Thus, we obtained a set of 4 values (Delta, Theta, Alpha, and Beta) of relative power with a time resolution of 6 s. 2.2.2. SO definition NREM sleep is composed of 4 stages which range from light (stages 1 and 2) to deep sleep (stages 3 and 4). According to the standard criteria, SO is defined as the first of two consecutive NREM stage 1 epochs (epoch duration= 30 s period), or the first epoch of any other sleep stage. We found this definition confusing in patients who had difficulties falling asleep, and were oscillating between wakefulness and sleep. Since we were interested in investigating the SO process at the point where sleep became steady, we looked for a reference point that will mark the onset of the transition to unequivocal sleep. We found that standard definition of SO coincided best with an alpha power decrease below 28% of total EEG energy. We used this definition as a reference point for the moment of SO with an additional requirement that alpha power remains below this level for at least 5 min, to ascertain that sleep was steady. Although we aimed at studying the SO process, we had to use a common point of reference in order to be able to compare the process between subjects. Thus, the definition we used for SO point, could be easily used as an objective measure. All recorded signals were analyzed relative to SO reference point. The period selected for analysis included 19 min time intervals centered around this point. 2.2.3. Respiration Respiratory effort signal, recorded from the thorax, was analyzed by a Short-Time-Fourier-Transform with a 30 s time window and 27.5 s of overlap, to obtain a spectrum every 2.5 s. Mean frequency of respiration was calculated by a weighted integral over the range 0.1–0.4 Hz. 2.2.4. HRV analysis R waves were automatically detected from the ECG signal, and their occurrences as a function of time composed
Table 1 Average values (and standard deviation) before and after SO for the 3 groups of subjects Parameter
Normals Before SO
After SO
Before SO
After SO
Before SO
After SO
Resp F Resp F var EMG EMG var RRI RRI var VLF LF HF LF/HF
1.029 (0.125) 1.919 (1.000) ⁎⁎⁎ 2.163 (1.581) ⁎ 2.990 (2.183) ⁎⁎⁎ 0.939 (0.034) ⁎⁎⁎ 1.158 (0.342) ⁎ 2.586 (1.777) ⁎⁎⁎ 1.719 (1.099) ⁎ 1.027 (0.828) 2.053 (1.358) ⁎⁎
1.000 (0.004) 1.024 (0.049) 1.001 (0.013) 0.993 (0.016) 0.999 (0.002) 0.988 (0.006) 1.016 (0.060) 0.987 (0.031) 0.990 (0.011) 1.048 (0.120)
0.992 (0.077) 1.579 (0.408) ⁎⁎⁎ 2.032 (1.002) ⁎⁎⁎ 2.929 (1.757) ⁎⁎⁎ 0.969 (0.041) ⁎⁎ 1.144 (0.341) 2.643 (1.610) ⁎⁎⁎ 1.461 (0.943) 0.982 (0.407) 1.755 (0.796) ⁎⁎
1.000 (0.003) 0.995 (0.026) 1.005 (0.019) 0.996 (0.014) 1.000 (0.003) 1.008 (0.016) 1.050 (0.092) 1.029 (0.068) 1.001 (0.017) 1.073 (0.088)
1.092 (0.084) ⁎⁎⁎ 1.956 (0.491) ⁎⁎⁎ 2.079 (1.222) ⁎⁎ 3.312 (2.496) ⁎⁎⁎ 0.956 (0.060) ⁎⁎ 1.385 (0.208) ⁎⁎⁎ 2.425 (1.169) ⁎⁎⁎ 2.029 (0.764) ⁎⁎⁎ 1.353 (0.587) ⁎ 1.643 (0.429) ⁎⁎⁎
1.000 (0.002) 0.999 (0.029) 1.002 (0.011) 0.997 (0.012) 1.000 (0.001) 0.991 (0.014) 1.012 (0.032) 0.984 (0.018) 0.988 (0.013) 1.001 (0.099)
⁎ p < 0.05. ⁎⁎ p < 0.01. ⁎⁎⁎ p < 0.005.
OSAS
VSD
20
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27
21
Fig. 2. Graphic presentation of the variability of: (1) respiration frequency in the upper frame, (2) EMG in mid frame, and (3) RRI in the bottom frame. Parameters averaged over all subjects (n = 34), minute-by-minute relative to SO time (as defined in the text). Averaging began 9 min before and ended 9 min after SO. Dark bars indicate the reference values that were used for comparisons to the other values (detailed explanation in the text). Asterisks mark the bars that had a significant different value (p < 0.05) in comparison to the reference values.
the RR interval series (RRI). RRI was interpolated to equally spaced samples, and its time–frequency decomposition was performed by a continuous wavelet algorithm (Toledo et al., 2003). This algorithm was specially developed to deal with non-stationary RRI series. It uses shorter time windows to estimate fluctuations of higher frequencies and wider windows for lower frequencies, thus achieving optimal time resolution for each frequency. The Power in 3 standard frequency bands was calculated: VLF—Very Low Frequency (0.005–0.04 Hz), LF—Low Frequency (0.04–0.15 Hz), HF—High Frequency (0.15–0.5 Hz). Due to large inter-subject variability, the amplitudes and powers of all parameters (except respiratory mean frequency) were calculated relative to a baseline. This baseline was calculated for each parameter as its average value during the 5 min between the 5th and the 9th minute after the SO reference point. Dividing the parameters by their corresponding baseline values normalized them and enabled comparisons and statistical analysis of different subjects. Finally, since some of the parameters showed a marked change in their variability before and after SO, we also calculated a moving standard deviation for those para-
meters, namely respiratory mean frequency, EMG amplitude, and RRI. 2.2.5. Statistical analysis This study addressed three main issues: First we searched for variables that changed significantly with SO, then we examined the pattern of these changes around SO; and finally we compared this pattern for the 3 groups of subjects. We applied a paired t-test on each parameter, to test whether its values changed significantly before and after SO; the results are summarized in Table 1. To examine more accurately the timing when the described changes had started, multiple t-test s were used and significant values were corrected according to the FDR method (Benjamini and Hochberg, 1995). Each t-test compared the values during the first 3 min of the selected 19 min (i.e. the values during the 9th–7th minutes before SO) to the values in each of the other 16 min afterwards. A graphic representation of these results appears in Figs. 1 and 2. Two-way (before, after SO) ANOVA was used to test for differences in the behavior of all the studied parameters between the 3 groups of patients.
Fig. 1. Parameters averaged over all subjects in the Normals group (n = 12). They are represented minute-by-minute relative to SO as defined in the text. The averaging started 9 min before and ended 9 min after SO. Dark bars indicate the reference values that were used for comparisons to the other values (detailed explanation in the text). Asterisks mark the bars that had a significant different value (p < 0.05) in comparison to the reference values.
22
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27
3. Results We start by reporting the results concerning straightforward physiological variables: Respiratory frequency, EMG and RRI. The respiratory frequency did not change significantly during SO, either in Normals or in OSAS groups, but it did change significantly in the VSD group. However the variability of respiratory frequency presented a very distinct trend, similar in all three groups. Thus, although breathing rate did not change significantly during SO, it became significantly more stable (less fluctuations) after SO. EMG exhibited a significant two-fold reduction in amplitude, and its variability declined by a factor of 3 after SO in all groups. RRI displayed a significant increase, representing a decline in HR during SO. Overall RRI variability decreased slightly during the same time, however this diminution was statistically significant only for Normals (p < 0.05) and VSD (p < 0.005). The most interesting findings in this study are related to the time–frequency analysis of RRI, especially to the power in VLF band. VLF power decreased 2.5 fold after SO, in all groups. The decrease was significant in all 3 groups (paired t-test between values before and after SO, p < 0.005). Applying the minuteby-minute multiple t-test we found that VLF power started to decrease 2 min before SO, and that this decrease was significant in comparison to its initial values. A less significant, although consistent, decrease was observed in the LF power, for the Normals (p < 0.05) and
VSD (p < 0.005) groups only. No significant change was detected in the HF power. However, the ratio of LF/HF power, showed a more significant decrease during SO, than either LF or HF considered separately—see Table 1. The minute-by-minute analysis showed that the significant change in parameters (except VLF power), occurred at SO or 1 min later. It is interesting to note that all parameters that measure variability in the time or frequency domains reached a local minimum 1–2 min after the SO reference point (Figs. 1 and 2). Next we addressed the possibility that the HRV parameters reflecting autonomic modulation behaved differently in the 3 groups during the transition from wakefulness to sleep. The results showed no significant difference among the three study groups during SO, in any of the parameters other than the respiratory frequency. This was the only parameter that showed a significant group and interaction effect (between groups p = 0.034, interaction effect p = 0.034). Although the behavior of VLF, LF, HF, and LF/HF was similar in all groups the level of autonomic activity differed significantly between groups. The sympathovagal balance reflected best the differences between groups, with Normals showing the lowest balance, OSAS being more sympathetically driven, and VSD displaying the highest sympathovagal balance (see Fig. 3). Most subjects (32 out 34) fell asleep without difficulty. A representative example of this pattern appears in Fig. 4. This subject fell asleep at epoch 25 according to standard criteria. He entered SWS gradually, starting from stage 1 to deeper
Fig. 3. The sympathovagal balance as reflected by LF/HF ratio in all groups of subjects around SO. The SO reference point is marked by the vertical line in the middle and the horizontal axis shows the time scale in minutes relative to SO. A decreasing trend towards parasympathetic dominance is common to all groups; however, each group had significantly different balance levels (p < 0.01). VSD subjects that complained about fatigue had the highest sympathovagal balance, OSAS patients were in the middle, and Normals with the lowest values.
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27
stages 2, 3 and 4, with no further sleep–wake fluctuations. The change in alpha power paralleled this process with a sharp decrease at SO. Delta power behaved as a mirror image
23
of alpha, sharply increasing at SO. The threshold we defined in alpha power falls within the same epoch as standard SO definition (vertical line). Note that the time–frequency
Fig. 4. Analysis results for a normal subject with smooth sleep onset. The abscissa is the time scale in epochs of 30s around the SO reference point. The vertical line at epoch 24.8 indicates the reference point, where Alpha power decreased below 28% of total power. From top to bottom the frames represent: (1) Standard sleep staging (Hypnogram) obtained according to R&K criteria. (2) The power in Alpha and Delta frequency bands extracted from C4–A1 EEG. A sharp decrease in Alpha power at SO is synchronous with a reciprocal strong increase in Delta power. (3) Mean respiration frequency extracted from the thorax effort belt. The slight increase in this parameter, seen in this graph, is not a representative example of respiration frequency behavior. The variability in Respiratory frequency (seen as fluctuations around the local average) decreased, and this trend was consistent for the whole group of patients. (4) Submental EMG amplitude as recorded; A decrease in the number and amplitude of muscle twitches can be observed (5) RRI series obtained from the ECG; RRI values increased gradually with SO, while its variability decreased. (6–8) Power in VLF, LF and HF frequency bands obtained by time–frequency analysis of RRI; All decreased during SO, however HF returned to its pre-SO level after few minutes. (9) LF/HF ratio.
24
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27
Fig. 5. Analysis results before and after SO for a normal subject that had trouble developing steady sleep. Steady sleep was reached only after epoch 56 (28 min). The frames appear in the same order as in Fig. 4. Note the fluctuating behavior of all parameters before SO, including EEG alpha and delta power, compared with the smooth behavior of the same parameters shown in Fig. 4. Note that the power of VLF decreases only before long periods of steady sleep, as seen around epoch 20 and after epoch 56.
parameters VLF and LF also decreased during the process, with VLF preceding SO. A different example of the correlation between these two parameters and the state of vigilance is presented in Fig. 5. It represents a case of fluctuating vigilance before sleep state stabilization in one of the subjects included in the study, who complained of difficulties to fall asleep. It is of interest to note
that the wake–sleep swing that appears in the hypnogram is accompanied by similar oscillations in VLF, LF and HF power. 4. Discussion The transition from wakefulness to sleep is an amalgam of cognitive, behavioral and physiological changes. We tried to
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27
zoom in on some of the physiological changes that are part of this amalgam, mainly the autonomic nervous changes. An attempt was made to define the connections between these changes, as revealed by instantaneous HRV, and the somewhat arbitrary standard definition of sleep onset. All observed changes occur at slightly different timings around the defined moment of SO, leading to the necessary conclusion that falling asleep represents a process rather than an event occurring at a definite time point (Ogilvie, 2001). This understanding lead us to try and define, based on consistent physiologic changes during the wake–sleep transition, a point in time that correlated best with the gold standard SO. The rationale for the choice of the specific definition used in our study is apparent in the example presented in Fig. 5. Note that according to standard R&K criteria, SO occurred at epoch 16, i.e. 8 min after lights off. However the patient reached stable sleep only 20 min later, at epoch 56. The alpha power in the EEG signal displayed oscillations during this time interval, to reach stable levels below the 28% of total EEG power only after epoch 56. Thus, our selection of sleep onset timing, corresponds with the standard SO definition for patients who fall asleep uneventfully, with the additional benefit that it matches the development of stable, unequivocal sleep in subjects who have difficulties to initiate sleep. The fine correlation between our quantitative definition of sleep onset and the gold standard allowed us to look into physiologic changes around the sustained drop in alpha power below 28% of total EEG power as an excellent equivalent of SO point. The changes described in the Results section included discrete changes in RR interval, breathing pattern and muscle activity, as well as striking changes in the indirect autonomic indicators, namely power spectral components of HRV. The average RRI increased (heart rate decreased) during SO, confirming findings in other studies (Pivik and Busby, 1996; Zemaityte et al., 1984). However, there are some subtle differences between our findings and those in (Pivik and Busby, 1996). Pivik and Busby found no significant change in RRI after SO when compared to wakefulness, however they found that RRI differed significantly when the levels preceding NREM stage 1 were compared with the 15 s after this stage started. We found that RRI levels changed significantly 30 s after SO compared to the average RRI level during 9–7 min before SO. This slight difference in our findings may result from the fact that in the previous study the wakefulness definition was not delineated, and that different definitions of SO were used (although in both studies a smooth transition to stable sleep was sought after). Moreover the studies dealt with subjects of greatly different ages: adults in ours and preadolescents in the previous one. The findings concerning changes in respiration are also of interest. While mean respiratory frequency did not change, its variability decreased with SO, thus breathing became more stable after falling asleep. These results corroborate
25
with those of (Trinder et al., 1992) although technical limitations did not allow us to evaluate tidal volume, which decreased according to the mentioned study (Trinder et al., 1992). The change in EMG amplitude during the transition from relaxed wakefulness to unequivocal sleep is referred to qualitatively as minor or indistinguishable (Rechtschaffen and Kales, 1968). Our approach was quantitative and revealed a significant decrease both in the amplitude and in the variability of the EMG across the SO process. The decrease in variability can be attributed to the number and magnitude of muscle twitches occurring before sleep onset. Twitches and movements are common before falling asleep and become rare afterwards as displayed in the EMG panel in Figs. 4 and 5. Some increase in the measured EMG parameters is apparent from 3 to 1 min prior to SO, as shown in Figs. 1 and 2, and might be attributed to movements. The main findings of the present study concern the changes in autonomic function that occur around SO. These findings are the result of the analysis of RRI in the time– frequency domain. VLF power decreased significantly in all 3 groups, while HF power did not change significantly in either one. The LF power displayed a trend similar to VLF, however the decrease was significant only for Normals and VSD and not for OSAS patients. Thus the source of the change in heart rate variability is the decrease in VLF and LF power. Since VLF and LF power components reflect both sympathetic and parasympathetic modulation at the sinus node, and HF power reflects mainly parasympathetic activity, it is plausible that the decrease in LF and VLF power reflects a reduced activity of the sympathetic branch of the ANS. The different behavior of VLF and LF in the OSAS group leads to the conclusion that the sympathetic nervous system does not reduce its activity significantly with SO in these patients. This might be the result of an increased number of arousals caused by repetitive respiratory events during sleep. The behavior of the VLF during SO was interesting. It displayed the most significant change, among all the frequency domain parameters. Moreover, VLF began its decrease well before SO, certainly before significant EEG changes occurred. VLF power changes are attributable to thermoregulation (Fleisher et al., 1996; Shefi, 1997) and physical activity (Bernardi et al., 1996, 2000) among other factors. Thus, the behavior of VLF just before SO may represent both a reduction in sympathetic activity that derives from cessation of activities while lying supine, and a reduced thermoregulation in a very stable thermal environment in bed before falling asleep. It is known that HF power increases with respiration depth (tidal volume). Thus, stable mean respiratory rate and a decrease in tidal volume, which have been reported in the literature (Trinder et al., 1992), may cause a decrease in HF power. The lack of a significant change in HF power suggests a possible concealed increase in parasympathetic activity accompanying SO.
26
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27
The decrease in VLF and LF power along with an unchanged HF power after SO, support earlier findings of increased parasympathetic activity during sleep stages 3 and 4 (Baharav et al., 1995; Berlad et al., 1993; Van de Borne et al., 1994). Furthermore, the significant decrease in autonomic balance as reflected by LF/HF ratio also indicates a shift in sympathovagal balance towards an increased parasympathetic activity after SO. The fact that the subjects were supine long before lights off suggests that the changes in autonomic modulation around sleep onset are related to the process of transition from wakefulness to sleep and not to the change from upright to supine position. Although the autonomic indices behaved similarly in all subjects, the study uncovered increased sympathovagal balance in both groups with abnormal sleep studies, as compared to Normals. These results indicate that patients with OSAS have predominantly sympathetic autonomic balance during sleep as well as during wakefulness before sleep onset. Moreover, patients in the VSD group who presented with a main complaint of fatigue, had even higher sympathetic predominance. These results are supported by (Narkiewicz et al., 1999; Somers et al., 1995) in OSAS and by (Stewart, 2000) in fatigue patients. The transition from wakefulness to sleep is accompanied by a fluctuating decrease in the parameters that reflect variability (Respiratory frequency variability, EMG variability, RRI variability, VLF, LF, HF, and LF/HF), until a nadir is reached shortly after SO (see Figs. 1 and 2). Further fluctuations of reduced amplitude follow this point. This striking finding indicates that SO is a transitional process between two physiologically different states. The magnitude of physiologic measures (HR, EMG amplitude) decreases with the transition to sleep. These findings are most likely connected to the falling asleep process, since preliminary results dealing with SO after awakenings during the night reproduce the results described here and corroborate with the changes in cardiovascular function in young adults during sleep onset (Carrington et al., 2005). Our findings add a new perspective on the wake– sleep transition as a step wise change or resetting of autonomic function from a certain level before SO to a lower one afterwards. During this transition, both the average level and the amplitude of the fluctuations of HRV parameters decreased, and the autonomic function reached a minimum. However, wake–sleep transition is in fact a process that starts sometime before that point, with a decrease in very slow fluctuations in HR, representing a decreased thermoregulation (Fleisher et al., 1996; Shefi, 1997), lower vasomotion (Taylor, 1998) as well as less renin–angiotensin effect (Akselrod, 1995). This prologue is followed by a trend towards a decrease in sympathovagal balance sometime after SO.
References Akselrod, S., 1995. Basic principles of heart rate variability. In: Malik, M., Camm, A.J. (Eds.), Heart Rate Variability. Futura Publishing Company, New York, pp. 147–163. Baharav, A., Kotagal, S., Gibbons, V., Rubin, B.K., Pratt, G., Karin, J., Akselrod, S., 1995. Fluctuations in autonomic nervous activity during sleep displayed by power spectrum analysis of heart rate variability. Neurology 45 (6), 1183–1187. Benjamini, B., Hochberg, Y., 1995. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J.R. Stat. Soc. B. 57, 289–300. Berlad, I.I., Shlitner, A., Ben Haim, S., Lavie, P., 1993. Power spectrum analysis and heart rate variability in Stage 4 and REM sleep: evidence for state-specific changes in autonomic dominance. J. Sleep Res. 2 (2), 88–90. Bernardi, L., Valle, F., Coco, M., Calciati, A., Sleight, P., 1996. Physical activity influences heart rate variability and very-low-frequency components in Holter electrocardiograms. Cardiovasc. Res. 32 (2), 234–237. Bernardi, L., Wdowczyk-Szulc, J., Valenti, C., Castoldi, S., Passino, C., Spadacini, G., Sleight, P., 2000. Effects of controlled breathing, mental activity and mental stress with or without verbalization on heart rate variability. J. Am. Coll. Cardiol. 35 (6), 1462–1469. Berntson, G.G., Bigger Jr., J.T., Eckberg, D.L., Grossman, P., Kaufmann, P. G., Malik, M., Nagaraja, H.N., Porges, S.W., Saul, J.P., Stone, P.H., van der Molen, M.W., 1997. Heart rate variability: origins, methods, and interpretive caveats. Psychophysiology 34 (6), 623–648. Bonnet, M.H., Moore, S.E., 1982. The threshold of sleep: perception of sleep as a function of time asleep and auditory threshold. Sleep 5 (3), 267–276. Burgess, H.J., Trinder, J., Kim, Y., Luke, D., 1997. Sleep and circadian influences on cardiac autonomic nervous system activity. Am. J. Physiol. 273 (4 Pt 2), H1761–H1768. Burgess, H.J., Kleiman, J., Trinder, J., 1999a. Cardiac activity during sleep onset. Psychophysiology 36 (3), 298–306. Burgess, H.J., Trinder, J., Kim, Y., 1999b. Cardiac autonomic nervous system activity during presleep wakefulness and stage 2 NREM sleep. J. Sleep Res. 8 (2), 113–122. Carrington, M., Walsh, M., Stambas, T., Kleiman, J., Trinder, J., 2003. The influence of sleep onset on the diurnal variation in cardiac activity and cardiac control. J. Sleep Res. 12 (3), 213–221. Carrington, M.J., Barbieri, R., Colrain, I.M., Crowley, K.E., Kim, Y., Trinder, J., 2005. Changes in cardiovascular function during sleep onset in young adults. J. Appl. Physiol. 98 (2), 468–476. Fleisher, L.A., Frank, S.M., Sessler, D.I., Cheng, C.T., Matsukawa, T., Vannier, C.A., 1996. Thermoregulation and heart rate variability. Clin. Sci. 90 (2), 97–103. Foulkes, D., Vogel, G., 1965. Mental activity at sleep onset. J. Abnorm. Psychology 70, 231–243. Hornyak, M., Cejnar, M., Elam, M., Matousek, M., Wallin, B.G., 1991. Sympathetic muscle nerve activity during sleep in man. Brain 114 (Pt. 3), 1281–1295. Keselbrener, L., Akselrod, S., 1996. Selective discrete Fourier transform algorithm for time–frequency analysis: method and application on simulated and cardiovascular signals. IEEE Trans. Biomed. Eng. 43 (8), 789–802. Malliani, A., 1995. Association of heart rate variability components with physiological regulatory mechanisms. In: Malik, M., Camm, A.J. (Eds.), Heart Rate Variability. Futura Publishing Company, New York, pp. 173–188. Narkiewicz, K., Kato, M., Phillips, B.G., Pesek, C.A., Davison, D.E., Somers, V.K., 1999. Nocturnal continuous positive airway pressure decreases daytime sympathetic traffic in obstructive sleep apnea. Circulation 100 (23), 2332–2335. Ogilvie, R.D., 2001. The process of falling asleep. Sleep Med. Rev. 5 (3), 247–270.
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27 Ogilvie, R.D., Wilkinson, R.T., Allison, S., 1989. The detection of sleep onset: behavioral, physiological, and subjective convergence. Sleep 12 (5), 458–474. Parmeggiani, P.L., Morrison, A.R., 1990. Alterations in autonomic functions during sleep. In: Loewy, A.D., Spayer, K.M. (Eds.), Central Regulation of Autonomic Functions. Oxford University Press, Oxford, pp. 367–386. Pivik, R.T., Busby, K., 1996. Heart rate associated with sleep onset in preadolescents. J. Sleep Res. 5 (1), 33–36. Rechtschaffen, A., Kales, A., 1968. A Manual of Standardized Terminology, Techniques and Scoring System for Sleep Stages of Human Subjects. NIH Publication, vol. 204. US Government Printing Office, Washington, DC. Ref Type: Generic. Shefi, O., 1997. Analysis of the thermoregulation mechanism in Neonates and adults by physiological signal processing, Tel-Aviv University. Somers, V.K., Dyken, M.E., Mark, A.L., Abboud, F.M., 1993. Sympatheticnerve activity during sleep in normal subjects. N. Engl. J. Med. 328 (5), 303–307. Somers, V.K., Dyken, M.E., Clary, M.P., Abboud, F.M., 1995. Sympathetic neural mechanisms in obstructive sleep apnea. J. Clin. Invest. 96 (4), 1897–1904.
27
Stewart, J.M., 2000. Autonomic nervous system dysfunction in adolescents with postural orthostatic tachycardia syndrome and chronic fatigue syndrome is characterized by attenuated vagal baroreflex and potentiated sympathetic vasomotion. Pediatr. Res. 48 (2), 218–226. Taylor, J.A., Carr, D.L., Myers, C.W., Eckberg, D.L., 1998. Mechanisms underlying very-low-frequency RR-interval oscillations in humans. Circulation 98 (6), 547–555. Toledo, E., Gurevitz, O., Hod, H., Eldar, M., Akselrod, S., 2003. Wavelet analysis of instantaneous heart rate: a study of autonomic control during thrombolysis. Am. J. Physiol., Regul. Integr. Comp. Physiol. 284 (4), R1079–R1091. Trinder, J., Whitworth, F., Kay, A., Wilkin, P., 1992. Respiratory instability during sleep onset. J. Appl. Physiol. 73 (6), 2462–2469. Van de Borne, P., Nguyen, H., Biston, P., Linkowski, P., Degaute, J.P., 1994. Effects of wake and sleep stages on the 24-h autonomic control of blood pressure and heart rate in recumbent men. Am. J. Physiol. 266 (2 Pt 2), H548–H554. Vogel, G., Foulkes, D., Trosman, H., 1966. Ego functions and dreaming during sleep onset. Arch. Gen. Psychiatry 14 (3), 238–248. Zemaityte, D., Varoneckas, G., Sokolov, E., 1984. Heart rhythm control during sleep. Psychophysiology 21 (3), 279–289.
Autonomic changes during wake–sleep transition: A heart rate variability based approach Zvi Shinar a,⁎, Solange Akselrod a , Yaron Dagan b , Armanda Baharav a a b
Abramson Center for Medical Physics, Tel Aviv University, Tel-Aviv, Israel Sleep and Fatigue Institute, SHEBA Medical Center, Tel Hashomer, Israel
Received 8 November 2005; received in revised form 3 April 2006; accepted 28 April 2006
Abstract Autonomic function during sleep and wakefulness has been extensively investigated, however information concerning autonomic changes during the wake to sleep transition is scarce. The objective of the present study was to non-invasively characterize autonomic function and additional physiologic changes during sleep onset in normal and abnormal sleep. The estimation of autonomic function was based on time– frequency analysis of the RR interval series, using the power components in the very-low-frequency range (0.005–0.04 Hz), low-frequency (0.04–0.15 Hz), and high-frequency range (0.15–0.5 Hz). The ratio of low to high frequency power represented the sympathovagal balance. Thirty-four subjects who underwent whole night polysomnography were divided into 3 groups according to their complaints and study results: normal subjects, apneic patients (OSAS), and subjects with various sleep disorders (VSD). The results indicated a significant increase in RR interval during sleep onset, although its variability decreased; respiratory rate did not change, yet respiration became more stable; EMG amplitude and its variability decreased with sleep onset. Very-low-frequency power started to decrease significantly 2 min before sleep onset in all groups; low-frequency power decreased and high-frequency power did not change significantly in all groups, accordingly their ratio decreased and reflected a shift towards parasympathetic predominance. Although autonomic function displayed similar behavior in all subjects, OSAS and VSD patients presented a higher sympathovagal balance reflecting enhanced sympathetic predominance in those groups compared to normal subjects, both before and after sleep onset. All parameters reached a nadir at a defined time point during the process of falling asleep. We conclude that the wake–sleep transition period represents a transitional process between two physiologically different states; this transition starts with a decrease in the very slow oscillations in heart rate that anticipates a step-change resetting of autonomic function, followed by a decrease in sympathovagal balance towards the end of the process. © 2006 Elsevier B.V. All rights reserved. Keywords: Autonomic nervous system; HRV; Wavelet; Sleep onset; Electrocardiogram
1. Introduction Sleep Onset (SO) is a transition process between two different physiological states: wakefulness and sleep, usually Non Rapid Eye Movement (NREM) sleep. The central regulation of autonomic function differs greatly between NREM and Rapid Eye Movement (REM) sleep (Parmeg⁎ Corresponding author. Medical Physics, School of Physics and Astronomy, Tel Aviv University, Tel Aviv 69978, Israel. Tel.: +972 3 6408669 or 6408204; fax: +972 3 6406237. E-mail address: [email protected] (Z. Shinar). 1566-0702/$ - see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.autneu.2006.04.006
giani and Morrison, 1990; Somers et al., 1993), yet little is known about the changes that occur during SO specifically. The definition of SO is rather vague as it depends on the physiologic or behavioral parameter under specific scrutiny (Ogilvie, 2001). Thus the standard definition (Rechtschaffen and Kales, 1968) of SO is based on consistent and reproducible changes in EEG, EOG, and EMG, and it usually allows to detect the transition from wakefulness to sleep within seconds. The subjective perception of falling asleep does not always coincide with the timing of the event as estimated by the standard definition. The entire process of falling asleep is further complicated by the fact that SO does not occur all at
18
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27
once (Ogilvie, 2001; Ogilvie et al., 1989) and some fluctuations in vigilance may occur before stable, unequivocal sleep develops. The complexity of the normal transition from quiet wakefulness to NREM sleep can be further demonstrated by the study of behavioral parameters during SO, such as the performance of simple behavioral tasks, memory tasks, or the response to auditory or visual stimuli (Bonnet and Moore, 1982; Foulkes and Vogel, 1965; Vogel et al., 1966). Alpha frequency EEG is typical for wakefulness with closed eyes, and in most subjects it appears before SO. The power in Delta frequency band increases during light sleep (NREM stage 2) as an intermediate state between wake and deep sleep. Delta frequency EEG is dominant during the deep stages of NREM sleep. The direct investigation of autonomic nervous function is invasive, leading by itself to unwanted changes and reactions. Thus direct studies of autonomic activity are usually avoided due to their relative complexity and invasiveness. Fortunately enough, important information can be inferred from the response of end organs in general, and from the heart rate variability (HRV) in particular. Thus standard methods of HRV in the time domain (Fast Fourier Transform or Autoregression) are widely accepted non-invasive tools, that allow to obtain important information about central autonomic regulation (Akselrod, 1995; Berntson et al., 1997;Malliani, 1995). The obtained HRV-power spectrum typically includes: a low frequency (LF) power component between (0.04– 0.15 Hz) which (in supine position) reflects both sympathetic and parasympathetic effects on the sinus node, and a distinct high frequency (HF) power component, around the respiratory frequency (0.15–0.5 Hz) which is mediated mainly by parasympathetic activity. Since LF includes information about both branches of the ANS, an estimate of autonomic or sympathovagal balance has been defined as the LF/HF ratio (Malliani, 1995). Slower HRV fluctuations in the range (0.005–0.04 Hz) known as Very Low Frequency (VLF) have been attributed to vasomotion, thermoregulation, and renin– angiotensin effects (Akselrod, 1995; Fleisher et al., 1996; Shefi, 1997; Taylor et al., 1998), and may be influenced, at least in part, by mental or physical activity (Bernardi et al., 1996, 2000). Thus VLF, LF, HF, and the LF/HF ratio provide a reliable estimate of cardiac autonomic control. One of the main disadvantages of power spectrum analysis resides in its limitation to steady state conditions. Timedependent spectral decomposition of beat-to-beat variability in HR overcomes this stationarity limitation (Keselbrener and Akselrod, 1996; Toledo et al., 2003), and can be applied in non-steady state conditions encountered frequently in physiologic studies. This approach is necessary for the evaluation of the transient autonomic changes that accompany SO. Several direct measurements (Hornyak et al., 1991) based on muscle sympathetic nerve activity and many indirect (Baharav et al., 1995; Berlad et al., 1993; Van de Borne et al., 1994) studies based on instantaneous heart rate variability as an estimate of autonomic activity addressed ANS function during sleep. These studies indicated that parasympathetic
activity increases progressively after SO and sympathetic nervous system activity decreases after falling asleep. These trends continued with deepening NREM sleep. However, only a small number of studies dealt specifically with ANS function during SO. Most of these studies were based on a series of experiments specifically designed to identify circadian effects (Burgess et al., 1997, 1999a,b; Carrington et al., 2003) concerning the influence of sleep on measures of cardiac autonomic activity. They indicated that blood pressure decreased and sympathovagal balance shifted towards increased vagal activity, in close association with SO. Moreover respiratory sinus arrhythmia, as a measure of parasympathetic activity, showed a 24-h rhythm independent of sleep, whereas pre-ejection period as a measure of sympathetic activity only showed a 24-h rhythm if sleep occurred. These findings suggest that parasympathetic nervous system activity was under strong circadian influence, whereas the sympathetic nervous system activity was mostly influenced by the sleep system (Burgess et al., 1997). More recent results indicated that the shift towards vagal dominance was primarily a function of sleep, whereas the change in sympathetic activity at sleep onset reflected circadian influence (Carrington et al., 2003). Although the results of different studies aimed to uncover circadian and sleep related influences on autonomic function around wake–sleep transitions are not in total agreement, there is no doubt that important changes in autonomic activity occur around sleep onset. The objective of the present study was to characterize the wake–sleep transition process, with the main focus on autonomic function changes as reflected by the instantaneous HRV. We estimated the behavior of additional physiologic variables (HR, respiration, muscle activity, and cerebral activity as reflected in the electroencephalogram) around SO. We assumed that falling asleep presents a consistent pattern of changes in these variables, that are common to normal and abnormal sleep. 2. Materials and methods 2.1. Data acquisition Thirty-four patients referred to a sleep laboratory underwent whole night sleep studies. All subjects gave their informed consent to participate in the study and answered a sleep questionnaire. Studies included whole night polysomnography with digitization and recording of the following channels: 4 EEG channels (two central and two occipital leads), and 2 EOG channels sampled at 100 Hz with low pass filter at 35 Hz, chin EMG, and right/left Anterior Tibialis (EMG)-sampled at 100 Hz low pass filtered at 45 Hz; ECG channel sampled at 200 Hz, with low pass at 100 Hz and notch filter at 50 Hz; abdominal and chest respiratory effort, nasal and oral airflow, end tidal CO2, oxygen saturation and pulse wave, all sampled at 10 Hz, and low pass filtered at 5 Hz. Subjects were asked to lie down 5 min before lights off.
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27
The polysomnograms were manually scored, by an expert, according to Rechtschaffen and Kales (R&K) criteria (Rechtschaffen and Kales, 1968), and were screened for any abnormalities. Following the scoring procedure, subjects were classified in 3 groups: 1. Normal patients (Normals)—No sleep abnormality was diagnosed after the sleep study in otherwise healthy subjects referred to the sleep clinic for mild snoring or some vague sleep quality or daytime fatigue complaints— 12 subjects (6 males): average ± Standard Deviation for age: 28 ± 16 years, sleep latency: 11 ± 8 min, sleep efficiency: 89 ± 12%. 2. Patients with obstructive sleep apnea (OSAS)—11 subjects (8 males): average ± Standard Deviation for age: 40 ± 9 years, sleep latency: 13 ± 9 min, and sleep efficiency: 86 ± 9%, average RDI 33.3 ± 19.6. 3. Patients with various sleep disorders (VSD)—11 subjects (8 males): average ± Standard Deviation for age: 37 ± 18 years, sleep latency: 10 ± 7 min, and sleep efficiency: 87 ± 8%. Subjects in this group were referred to the sleep laboratory with a main complaint of fatigue. Seven of them had a background of a psychiatric disease (obsessive– compulsive disorder, or post-traumatic stress disorder), two a history of asthma, and the other two were otherwise healthy; however their sleep study revealed periodic leg movement syndrome in one (PLM index 6.4), and central apnea (RDI 28.8) in the other. None had OSAS. No subject received any direct sympathetic agonist therapy, one patient in the OSAS group and two in the VSD received beta-blockers. 2.2. Data analysis 2.2.1. EEG signals The power of the EEG in the frequency bands of Delta (0.5–4 Hz), Theta (4–7 Hz), Alpha (7.5–14), and Beta (14– 35 Hz) was calculated for segments of 6 s. The power in each
19
frequency band was normalized by the total power in each segment. Thus, we obtained a set of 4 values (Delta, Theta, Alpha, and Beta) of relative power with a time resolution of 6 s. 2.2.2. SO definition NREM sleep is composed of 4 stages which range from light (stages 1 and 2) to deep sleep (stages 3 and 4). According to the standard criteria, SO is defined as the first of two consecutive NREM stage 1 epochs (epoch duration= 30 s period), or the first epoch of any other sleep stage. We found this definition confusing in patients who had difficulties falling asleep, and were oscillating between wakefulness and sleep. Since we were interested in investigating the SO process at the point where sleep became steady, we looked for a reference point that will mark the onset of the transition to unequivocal sleep. We found that standard definition of SO coincided best with an alpha power decrease below 28% of total EEG energy. We used this definition as a reference point for the moment of SO with an additional requirement that alpha power remains below this level for at least 5 min, to ascertain that sleep was steady. Although we aimed at studying the SO process, we had to use a common point of reference in order to be able to compare the process between subjects. Thus, the definition we used for SO point, could be easily used as an objective measure. All recorded signals were analyzed relative to SO reference point. The period selected for analysis included 19 min time intervals centered around this point. 2.2.3. Respiration Respiratory effort signal, recorded from the thorax, was analyzed by a Short-Time-Fourier-Transform with a 30 s time window and 27.5 s of overlap, to obtain a spectrum every 2.5 s. Mean frequency of respiration was calculated by a weighted integral over the range 0.1–0.4 Hz. 2.2.4. HRV analysis R waves were automatically detected from the ECG signal, and their occurrences as a function of time composed
Table 1 Average values (and standard deviation) before and after SO for the 3 groups of subjects Parameter
Normals Before SO
After SO
Before SO
After SO
Before SO
After SO
Resp F Resp F var EMG EMG var RRI RRI var VLF LF HF LF/HF
1.029 (0.125) 1.919 (1.000) ⁎⁎⁎ 2.163 (1.581) ⁎ 2.990 (2.183) ⁎⁎⁎ 0.939 (0.034) ⁎⁎⁎ 1.158 (0.342) ⁎ 2.586 (1.777) ⁎⁎⁎ 1.719 (1.099) ⁎ 1.027 (0.828) 2.053 (1.358) ⁎⁎
1.000 (0.004) 1.024 (0.049) 1.001 (0.013) 0.993 (0.016) 0.999 (0.002) 0.988 (0.006) 1.016 (0.060) 0.987 (0.031) 0.990 (0.011) 1.048 (0.120)
0.992 (0.077) 1.579 (0.408) ⁎⁎⁎ 2.032 (1.002) ⁎⁎⁎ 2.929 (1.757) ⁎⁎⁎ 0.969 (0.041) ⁎⁎ 1.144 (0.341) 2.643 (1.610) ⁎⁎⁎ 1.461 (0.943) 0.982 (0.407) 1.755 (0.796) ⁎⁎
1.000 (0.003) 0.995 (0.026) 1.005 (0.019) 0.996 (0.014) 1.000 (0.003) 1.008 (0.016) 1.050 (0.092) 1.029 (0.068) 1.001 (0.017) 1.073 (0.088)
1.092 (0.084) ⁎⁎⁎ 1.956 (0.491) ⁎⁎⁎ 2.079 (1.222) ⁎⁎ 3.312 (2.496) ⁎⁎⁎ 0.956 (0.060) ⁎⁎ 1.385 (0.208) ⁎⁎⁎ 2.425 (1.169) ⁎⁎⁎ 2.029 (0.764) ⁎⁎⁎ 1.353 (0.587) ⁎ 1.643 (0.429) ⁎⁎⁎
1.000 (0.002) 0.999 (0.029) 1.002 (0.011) 0.997 (0.012) 1.000 (0.001) 0.991 (0.014) 1.012 (0.032) 0.984 (0.018) 0.988 (0.013) 1.001 (0.099)
⁎ p < 0.05. ⁎⁎ p < 0.01. ⁎⁎⁎ p < 0.005.
OSAS
VSD
20
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27
21
Fig. 2. Graphic presentation of the variability of: (1) respiration frequency in the upper frame, (2) EMG in mid frame, and (3) RRI in the bottom frame. Parameters averaged over all subjects (n = 34), minute-by-minute relative to SO time (as defined in the text). Averaging began 9 min before and ended 9 min after SO. Dark bars indicate the reference values that were used for comparisons to the other values (detailed explanation in the text). Asterisks mark the bars that had a significant different value (p < 0.05) in comparison to the reference values.
the RR interval series (RRI). RRI was interpolated to equally spaced samples, and its time–frequency decomposition was performed by a continuous wavelet algorithm (Toledo et al., 2003). This algorithm was specially developed to deal with non-stationary RRI series. It uses shorter time windows to estimate fluctuations of higher frequencies and wider windows for lower frequencies, thus achieving optimal time resolution for each frequency. The Power in 3 standard frequency bands was calculated: VLF—Very Low Frequency (0.005–0.04 Hz), LF—Low Frequency (0.04–0.15 Hz), HF—High Frequency (0.15–0.5 Hz). Due to large inter-subject variability, the amplitudes and powers of all parameters (except respiratory mean frequency) were calculated relative to a baseline. This baseline was calculated for each parameter as its average value during the 5 min between the 5th and the 9th minute after the SO reference point. Dividing the parameters by their corresponding baseline values normalized them and enabled comparisons and statistical analysis of different subjects. Finally, since some of the parameters showed a marked change in their variability before and after SO, we also calculated a moving standard deviation for those para-
meters, namely respiratory mean frequency, EMG amplitude, and RRI. 2.2.5. Statistical analysis This study addressed three main issues: First we searched for variables that changed significantly with SO, then we examined the pattern of these changes around SO; and finally we compared this pattern for the 3 groups of subjects. We applied a paired t-test on each parameter, to test whether its values changed significantly before and after SO; the results are summarized in Table 1. To examine more accurately the timing when the described changes had started, multiple t-test s were used and significant values were corrected according to the FDR method (Benjamini and Hochberg, 1995). Each t-test compared the values during the first 3 min of the selected 19 min (i.e. the values during the 9th–7th minutes before SO) to the values in each of the other 16 min afterwards. A graphic representation of these results appears in Figs. 1 and 2. Two-way (before, after SO) ANOVA was used to test for differences in the behavior of all the studied parameters between the 3 groups of patients.
Fig. 1. Parameters averaged over all subjects in the Normals group (n = 12). They are represented minute-by-minute relative to SO as defined in the text. The averaging started 9 min before and ended 9 min after SO. Dark bars indicate the reference values that were used for comparisons to the other values (detailed explanation in the text). Asterisks mark the bars that had a significant different value (p < 0.05) in comparison to the reference values.
22
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27
3. Results We start by reporting the results concerning straightforward physiological variables: Respiratory frequency, EMG and RRI. The respiratory frequency did not change significantly during SO, either in Normals or in OSAS groups, but it did change significantly in the VSD group. However the variability of respiratory frequency presented a very distinct trend, similar in all three groups. Thus, although breathing rate did not change significantly during SO, it became significantly more stable (less fluctuations) after SO. EMG exhibited a significant two-fold reduction in amplitude, and its variability declined by a factor of 3 after SO in all groups. RRI displayed a significant increase, representing a decline in HR during SO. Overall RRI variability decreased slightly during the same time, however this diminution was statistically significant only for Normals (p < 0.05) and VSD (p < 0.005). The most interesting findings in this study are related to the time–frequency analysis of RRI, especially to the power in VLF band. VLF power decreased 2.5 fold after SO, in all groups. The decrease was significant in all 3 groups (paired t-test between values before and after SO, p < 0.005). Applying the minuteby-minute multiple t-test we found that VLF power started to decrease 2 min before SO, and that this decrease was significant in comparison to its initial values. A less significant, although consistent, decrease was observed in the LF power, for the Normals (p < 0.05) and
VSD (p < 0.005) groups only. No significant change was detected in the HF power. However, the ratio of LF/HF power, showed a more significant decrease during SO, than either LF or HF considered separately—see Table 1. The minute-by-minute analysis showed that the significant change in parameters (except VLF power), occurred at SO or 1 min later. It is interesting to note that all parameters that measure variability in the time or frequency domains reached a local minimum 1–2 min after the SO reference point (Figs. 1 and 2). Next we addressed the possibility that the HRV parameters reflecting autonomic modulation behaved differently in the 3 groups during the transition from wakefulness to sleep. The results showed no significant difference among the three study groups during SO, in any of the parameters other than the respiratory frequency. This was the only parameter that showed a significant group and interaction effect (between groups p = 0.034, interaction effect p = 0.034). Although the behavior of VLF, LF, HF, and LF/HF was similar in all groups the level of autonomic activity differed significantly between groups. The sympathovagal balance reflected best the differences between groups, with Normals showing the lowest balance, OSAS being more sympathetically driven, and VSD displaying the highest sympathovagal balance (see Fig. 3). Most subjects (32 out 34) fell asleep without difficulty. A representative example of this pattern appears in Fig. 4. This subject fell asleep at epoch 25 according to standard criteria. He entered SWS gradually, starting from stage 1 to deeper
Fig. 3. The sympathovagal balance as reflected by LF/HF ratio in all groups of subjects around SO. The SO reference point is marked by the vertical line in the middle and the horizontal axis shows the time scale in minutes relative to SO. A decreasing trend towards parasympathetic dominance is common to all groups; however, each group had significantly different balance levels (p < 0.01). VSD subjects that complained about fatigue had the highest sympathovagal balance, OSAS patients were in the middle, and Normals with the lowest values.
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27
stages 2, 3 and 4, with no further sleep–wake fluctuations. The change in alpha power paralleled this process with a sharp decrease at SO. Delta power behaved as a mirror image
23
of alpha, sharply increasing at SO. The threshold we defined in alpha power falls within the same epoch as standard SO definition (vertical line). Note that the time–frequency
Fig. 4. Analysis results for a normal subject with smooth sleep onset. The abscissa is the time scale in epochs of 30s around the SO reference point. The vertical line at epoch 24.8 indicates the reference point, where Alpha power decreased below 28% of total power. From top to bottom the frames represent: (1) Standard sleep staging (Hypnogram) obtained according to R&K criteria. (2) The power in Alpha and Delta frequency bands extracted from C4–A1 EEG. A sharp decrease in Alpha power at SO is synchronous with a reciprocal strong increase in Delta power. (3) Mean respiration frequency extracted from the thorax effort belt. The slight increase in this parameter, seen in this graph, is not a representative example of respiration frequency behavior. The variability in Respiratory frequency (seen as fluctuations around the local average) decreased, and this trend was consistent for the whole group of patients. (4) Submental EMG amplitude as recorded; A decrease in the number and amplitude of muscle twitches can be observed (5) RRI series obtained from the ECG; RRI values increased gradually with SO, while its variability decreased. (6–8) Power in VLF, LF and HF frequency bands obtained by time–frequency analysis of RRI; All decreased during SO, however HF returned to its pre-SO level after few minutes. (9) LF/HF ratio.
24
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27
Fig. 5. Analysis results before and after SO for a normal subject that had trouble developing steady sleep. Steady sleep was reached only after epoch 56 (28 min). The frames appear in the same order as in Fig. 4. Note the fluctuating behavior of all parameters before SO, including EEG alpha and delta power, compared with the smooth behavior of the same parameters shown in Fig. 4. Note that the power of VLF decreases only before long periods of steady sleep, as seen around epoch 20 and after epoch 56.
parameters VLF and LF also decreased during the process, with VLF preceding SO. A different example of the correlation between these two parameters and the state of vigilance is presented in Fig. 5. It represents a case of fluctuating vigilance before sleep state stabilization in one of the subjects included in the study, who complained of difficulties to fall asleep. It is of interest to note
that the wake–sleep swing that appears in the hypnogram is accompanied by similar oscillations in VLF, LF and HF power. 4. Discussion The transition from wakefulness to sleep is an amalgam of cognitive, behavioral and physiological changes. We tried to
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27
zoom in on some of the physiological changes that are part of this amalgam, mainly the autonomic nervous changes. An attempt was made to define the connections between these changes, as revealed by instantaneous HRV, and the somewhat arbitrary standard definition of sleep onset. All observed changes occur at slightly different timings around the defined moment of SO, leading to the necessary conclusion that falling asleep represents a process rather than an event occurring at a definite time point (Ogilvie, 2001). This understanding lead us to try and define, based on consistent physiologic changes during the wake–sleep transition, a point in time that correlated best with the gold standard SO. The rationale for the choice of the specific definition used in our study is apparent in the example presented in Fig. 5. Note that according to standard R&K criteria, SO occurred at epoch 16, i.e. 8 min after lights off. However the patient reached stable sleep only 20 min later, at epoch 56. The alpha power in the EEG signal displayed oscillations during this time interval, to reach stable levels below the 28% of total EEG power only after epoch 56. Thus, our selection of sleep onset timing, corresponds with the standard SO definition for patients who fall asleep uneventfully, with the additional benefit that it matches the development of stable, unequivocal sleep in subjects who have difficulties to initiate sleep. The fine correlation between our quantitative definition of sleep onset and the gold standard allowed us to look into physiologic changes around the sustained drop in alpha power below 28% of total EEG power as an excellent equivalent of SO point. The changes described in the Results section included discrete changes in RR interval, breathing pattern and muscle activity, as well as striking changes in the indirect autonomic indicators, namely power spectral components of HRV. The average RRI increased (heart rate decreased) during SO, confirming findings in other studies (Pivik and Busby, 1996; Zemaityte et al., 1984). However, there are some subtle differences between our findings and those in (Pivik and Busby, 1996). Pivik and Busby found no significant change in RRI after SO when compared to wakefulness, however they found that RRI differed significantly when the levels preceding NREM stage 1 were compared with the 15 s after this stage started. We found that RRI levels changed significantly 30 s after SO compared to the average RRI level during 9–7 min before SO. This slight difference in our findings may result from the fact that in the previous study the wakefulness definition was not delineated, and that different definitions of SO were used (although in both studies a smooth transition to stable sleep was sought after). Moreover the studies dealt with subjects of greatly different ages: adults in ours and preadolescents in the previous one. The findings concerning changes in respiration are also of interest. While mean respiratory frequency did not change, its variability decreased with SO, thus breathing became more stable after falling asleep. These results corroborate
25
with those of (Trinder et al., 1992) although technical limitations did not allow us to evaluate tidal volume, which decreased according to the mentioned study (Trinder et al., 1992). The change in EMG amplitude during the transition from relaxed wakefulness to unequivocal sleep is referred to qualitatively as minor or indistinguishable (Rechtschaffen and Kales, 1968). Our approach was quantitative and revealed a significant decrease both in the amplitude and in the variability of the EMG across the SO process. The decrease in variability can be attributed to the number and magnitude of muscle twitches occurring before sleep onset. Twitches and movements are common before falling asleep and become rare afterwards as displayed in the EMG panel in Figs. 4 and 5. Some increase in the measured EMG parameters is apparent from 3 to 1 min prior to SO, as shown in Figs. 1 and 2, and might be attributed to movements. The main findings of the present study concern the changes in autonomic function that occur around SO. These findings are the result of the analysis of RRI in the time– frequency domain. VLF power decreased significantly in all 3 groups, while HF power did not change significantly in either one. The LF power displayed a trend similar to VLF, however the decrease was significant only for Normals and VSD and not for OSAS patients. Thus the source of the change in heart rate variability is the decrease in VLF and LF power. Since VLF and LF power components reflect both sympathetic and parasympathetic modulation at the sinus node, and HF power reflects mainly parasympathetic activity, it is plausible that the decrease in LF and VLF power reflects a reduced activity of the sympathetic branch of the ANS. The different behavior of VLF and LF in the OSAS group leads to the conclusion that the sympathetic nervous system does not reduce its activity significantly with SO in these patients. This might be the result of an increased number of arousals caused by repetitive respiratory events during sleep. The behavior of the VLF during SO was interesting. It displayed the most significant change, among all the frequency domain parameters. Moreover, VLF began its decrease well before SO, certainly before significant EEG changes occurred. VLF power changes are attributable to thermoregulation (Fleisher et al., 1996; Shefi, 1997) and physical activity (Bernardi et al., 1996, 2000) among other factors. Thus, the behavior of VLF just before SO may represent both a reduction in sympathetic activity that derives from cessation of activities while lying supine, and a reduced thermoregulation in a very stable thermal environment in bed before falling asleep. It is known that HF power increases with respiration depth (tidal volume). Thus, stable mean respiratory rate and a decrease in tidal volume, which have been reported in the literature (Trinder et al., 1992), may cause a decrease in HF power. The lack of a significant change in HF power suggests a possible concealed increase in parasympathetic activity accompanying SO.
26
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27
The decrease in VLF and LF power along with an unchanged HF power after SO, support earlier findings of increased parasympathetic activity during sleep stages 3 and 4 (Baharav et al., 1995; Berlad et al., 1993; Van de Borne et al., 1994). Furthermore, the significant decrease in autonomic balance as reflected by LF/HF ratio also indicates a shift in sympathovagal balance towards an increased parasympathetic activity after SO. The fact that the subjects were supine long before lights off suggests that the changes in autonomic modulation around sleep onset are related to the process of transition from wakefulness to sleep and not to the change from upright to supine position. Although the autonomic indices behaved similarly in all subjects, the study uncovered increased sympathovagal balance in both groups with abnormal sleep studies, as compared to Normals. These results indicate that patients with OSAS have predominantly sympathetic autonomic balance during sleep as well as during wakefulness before sleep onset. Moreover, patients in the VSD group who presented with a main complaint of fatigue, had even higher sympathetic predominance. These results are supported by (Narkiewicz et al., 1999; Somers et al., 1995) in OSAS and by (Stewart, 2000) in fatigue patients. The transition from wakefulness to sleep is accompanied by a fluctuating decrease in the parameters that reflect variability (Respiratory frequency variability, EMG variability, RRI variability, VLF, LF, HF, and LF/HF), until a nadir is reached shortly after SO (see Figs. 1 and 2). Further fluctuations of reduced amplitude follow this point. This striking finding indicates that SO is a transitional process between two physiologically different states. The magnitude of physiologic measures (HR, EMG amplitude) decreases with the transition to sleep. These findings are most likely connected to the falling asleep process, since preliminary results dealing with SO after awakenings during the night reproduce the results described here and corroborate with the changes in cardiovascular function in young adults during sleep onset (Carrington et al., 2005). Our findings add a new perspective on the wake– sleep transition as a step wise change or resetting of autonomic function from a certain level before SO to a lower one afterwards. During this transition, both the average level and the amplitude of the fluctuations of HRV parameters decreased, and the autonomic function reached a minimum. However, wake–sleep transition is in fact a process that starts sometime before that point, with a decrease in very slow fluctuations in HR, representing a decreased thermoregulation (Fleisher et al., 1996; Shefi, 1997), lower vasomotion (Taylor, 1998) as well as less renin–angiotensin effect (Akselrod, 1995). This prologue is followed by a trend towards a decrease in sympathovagal balance sometime after SO.
References Akselrod, S., 1995. Basic principles of heart rate variability. In: Malik, M., Camm, A.J. (Eds.), Heart Rate Variability. Futura Publishing Company, New York, pp. 147–163. Baharav, A., Kotagal, S., Gibbons, V., Rubin, B.K., Pratt, G., Karin, J., Akselrod, S., 1995. Fluctuations in autonomic nervous activity during sleep displayed by power spectrum analysis of heart rate variability. Neurology 45 (6), 1183–1187. Benjamini, B., Hochberg, Y., 1995. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J.R. Stat. Soc. B. 57, 289–300. Berlad, I.I., Shlitner, A., Ben Haim, S., Lavie, P., 1993. Power spectrum analysis and heart rate variability in Stage 4 and REM sleep: evidence for state-specific changes in autonomic dominance. J. Sleep Res. 2 (2), 88–90. Bernardi, L., Valle, F., Coco, M., Calciati, A., Sleight, P., 1996. Physical activity influences heart rate variability and very-low-frequency components in Holter electrocardiograms. Cardiovasc. Res. 32 (2), 234–237. Bernardi, L., Wdowczyk-Szulc, J., Valenti, C., Castoldi, S., Passino, C., Spadacini, G., Sleight, P., 2000. Effects of controlled breathing, mental activity and mental stress with or without verbalization on heart rate variability. J. Am. Coll. Cardiol. 35 (6), 1462–1469. Berntson, G.G., Bigger Jr., J.T., Eckberg, D.L., Grossman, P., Kaufmann, P. G., Malik, M., Nagaraja, H.N., Porges, S.W., Saul, J.P., Stone, P.H., van der Molen, M.W., 1997. Heart rate variability: origins, methods, and interpretive caveats. Psychophysiology 34 (6), 623–648. Bonnet, M.H., Moore, S.E., 1982. The threshold of sleep: perception of sleep as a function of time asleep and auditory threshold. Sleep 5 (3), 267–276. Burgess, H.J., Trinder, J., Kim, Y., Luke, D., 1997. Sleep and circadian influences on cardiac autonomic nervous system activity. Am. J. Physiol. 273 (4 Pt 2), H1761–H1768. Burgess, H.J., Kleiman, J., Trinder, J., 1999a. Cardiac activity during sleep onset. Psychophysiology 36 (3), 298–306. Burgess, H.J., Trinder, J., Kim, Y., 1999b. Cardiac autonomic nervous system activity during presleep wakefulness and stage 2 NREM sleep. J. Sleep Res. 8 (2), 113–122. Carrington, M., Walsh, M., Stambas, T., Kleiman, J., Trinder, J., 2003. The influence of sleep onset on the diurnal variation in cardiac activity and cardiac control. J. Sleep Res. 12 (3), 213–221. Carrington, M.J., Barbieri, R., Colrain, I.M., Crowley, K.E., Kim, Y., Trinder, J., 2005. Changes in cardiovascular function during sleep onset in young adults. J. Appl. Physiol. 98 (2), 468–476. Fleisher, L.A., Frank, S.M., Sessler, D.I., Cheng, C.T., Matsukawa, T., Vannier, C.A., 1996. Thermoregulation and heart rate variability. Clin. Sci. 90 (2), 97–103. Foulkes, D., Vogel, G., 1965. Mental activity at sleep onset. J. Abnorm. Psychology 70, 231–243. Hornyak, M., Cejnar, M., Elam, M., Matousek, M., Wallin, B.G., 1991. Sympathetic muscle nerve activity during sleep in man. Brain 114 (Pt. 3), 1281–1295. Keselbrener, L., Akselrod, S., 1996. Selective discrete Fourier transform algorithm for time–frequency analysis: method and application on simulated and cardiovascular signals. IEEE Trans. Biomed. Eng. 43 (8), 789–802. Malliani, A., 1995. Association of heart rate variability components with physiological regulatory mechanisms. In: Malik, M., Camm, A.J. (Eds.), Heart Rate Variability. Futura Publishing Company, New York, pp. 173–188. Narkiewicz, K., Kato, M., Phillips, B.G., Pesek, C.A., Davison, D.E., Somers, V.K., 1999. Nocturnal continuous positive airway pressure decreases daytime sympathetic traffic in obstructive sleep apnea. Circulation 100 (23), 2332–2335. Ogilvie, R.D., 2001. The process of falling asleep. Sleep Med. Rev. 5 (3), 247–270.
Z. Shinar et al. / Autonomic Neuroscience: Basic and Clinical 130 (2006) 17–27 Ogilvie, R.D., Wilkinson, R.T., Allison, S., 1989. The detection of sleep onset: behavioral, physiological, and subjective convergence. Sleep 12 (5), 458–474. Parmeggiani, P.L., Morrison, A.R., 1990. Alterations in autonomic functions during sleep. In: Loewy, A.D., Spayer, K.M. (Eds.), Central Regulation of Autonomic Functions. Oxford University Press, Oxford, pp. 367–386. Pivik, R.T., Busby, K., 1996. Heart rate associated with sleep onset in preadolescents. J. Sleep Res. 5 (1), 33–36. Rechtschaffen, A., Kales, A., 1968. A Manual of Standardized Terminology, Techniques and Scoring System for Sleep Stages of Human Subjects. NIH Publication, vol. 204. US Government Printing Office, Washington, DC. Ref Type: Generic. Shefi, O., 1997. Analysis of the thermoregulation mechanism in Neonates and adults by physiological signal processing, Tel-Aviv University. Somers, V.K., Dyken, M.E., Mark, A.L., Abboud, F.M., 1993. Sympatheticnerve activity during sleep in normal subjects. N. Engl. J. Med. 328 (5), 303–307. Somers, V.K., Dyken, M.E., Clary, M.P., Abboud, F.M., 1995. Sympathetic neural mechanisms in obstructive sleep apnea. J. Clin. Invest. 96 (4), 1897–1904.
27
Stewart, J.M., 2000. Autonomic nervous system dysfunction in adolescents with postural orthostatic tachycardia syndrome and chronic fatigue syndrome is characterized by attenuated vagal baroreflex and potentiated sympathetic vasomotion. Pediatr. Res. 48 (2), 218–226. Taylor, J.A., Carr, D.L., Myers, C.W., Eckberg, D.L., 1998. Mechanisms underlying very-low-frequency RR-interval oscillations in humans. Circulation 98 (6), 547–555. Toledo, E., Gurevitz, O., Hod, H., Eldar, M., Akselrod, S., 2003. Wavelet analysis of instantaneous heart rate: a study of autonomic control during thrombolysis. Am. J. Physiol., Regul. Integr. Comp. Physiol. 284 (4), R1079–R1091. Trinder, J., Whitworth, F., Kay, A., Wilkin, P., 1992. Respiratory instability during sleep onset. J. Appl. Physiol. 73 (6), 2462–2469. Van de Borne, P., Nguyen, H., Biston, P., Linkowski, P., Degaute, J.P., 1994. Effects of wake and sleep stages on the 24-h autonomic control of blood pressure and heart rate in recumbent men. Am. J. Physiol. 266 (2 Pt 2), H548–H554. Vogel, G., Foulkes, D., Trosman, H., 1966. Ego functions and dreaming during sleep onset. Arch. Gen. Psychiatry 14 (3), 238–248. Zemaityte, D., Varoneckas, G., Sokolov, E., 1984. Heart rhythm control during sleep. Psychophysiology 21 (3), 279–289.